Non-rigid three-dimensional model classification method based on dual-channel convolutional neural network learning

The invention discloses a non-rigid three-dimensional model classification method based on double-channel convolutional neural network learning, and the method comprises the steps: firstly extractinga BoF feature vector of a three-dimensional model, and obtaining the intrinsic deep geometric features through a constructed BoF-CNN learning channel; secondly, based on an MVCNN (Multi-View Convolutional Neural Network), establishing a parallel 2D view CNN learning channel, and extracting an extrinsic depth view feature; further, connecting the view features with the geometric features, and constructing feature representation of the information image; and finally, further carrying out refining and weighted fusion through a neural network to generate differentiated depth feature representation, and achieving effective classification of the three-dimensional model is realized based on Softmax. The application range is wide, and the classification precision and efficiency are effectively improved. 本发明公开一种基于双通道卷积神经网络